I think ChatGPT in itself is a weak tool for all the reasons Marcus provides. But, the ways that it and other MLLs have already been integrated into new use cases is staggering. Whether using them to code whole new websites, short-cut the tedious working of designing websites, or coding simple apps or coding run-of-the-mill functionality— that’s all stuff that has emerged in less than 3 months! The sheer velocity is staggering.
For myself, various MLL search and citation tools have made much of my work markedly more efficient. For example, using a tool like perplexity.ai (whose same function will also be accessible through the new Bing platform) makes research for my own writing more efficient. Since the platform provides citations for all of its claims, they are subject to scrutiny.
Another similar tool that can generate responses to scientific questions (consensus.app) can provide primary sources in response to questions that would yield little in a Google Scholar search. For example, asking “Can rapamycin slow dementia or Alzheimer’s progression”, the tool responds with a sort of summary of its interpretation of the research —but, again, it also provides all of its source material with links to originals.
My point is this. ChatGPT is indeed a nascent version of MLL interfaces. But the potential for these models and their myriad applications is massive. Although I have been working in and around “disruptive” technology for over 15 years, I haven’t been this excited about AI or technology in general for a very long time. Autonomous vehicles seem to me much farther off as real, usable tools (and much less exciting) than MLL ‘s opening of whole new areas of work, inquiry, creativity and innovation.
Very interesting and insightful post by Gary. While I agree that ChatGPT 3 is error-prone, carries misinformation, does not offer source, it is worth remembering Clayton Christenson's Disruptive Innovation Model- While the incumbents are busy serving high quality products to their more demanding and profitable customer base, they tend to ignore the needs of others. With ChatGPTs ability to provide a cogent response (error-prone) , the customer base might tend to prioritize convenience(effortless) over effortful selection within links offered by the incumbent. In any case, arrival of ChatGPT is a watershed moment..
I think ChatGPT in itself is a weak tool for all the reasons Marcus provides. But, the ways that it and other MLLs have already been integrated into new use cases is staggering. Whether using them to code whole new websites, short-cut the tedious working of designing websites, or coding simple apps or coding run-of-the-mill functionality— that’s all stuff that has emerged in less than 3 months! The sheer velocity is staggering.
For myself, various MLL search and citation tools have made much of my work markedly more efficient. For example, using a tool like perplexity.ai (whose same function will also be accessible through the new Bing platform) makes research for my own writing more efficient. Since the platform provides citations for all of its claims, they are subject to scrutiny.
Another similar tool that can generate responses to scientific questions (consensus.app) can provide primary sources in response to questions that would yield little in a Google Scholar search. For example, asking “Can rapamycin slow dementia or Alzheimer’s progression”, the tool responds with a sort of summary of its interpretation of the research —but, again, it also provides all of its source material with links to originals.
My point is this. ChatGPT is indeed a nascent version of MLL interfaces. But the potential for these models and their myriad applications is massive. Although I have been working in and around “disruptive” technology for over 15 years, I haven’t been this excited about AI or technology in general for a very long time. Autonomous vehicles seem to me much farther off as real, usable tools (and much less exciting) than MLL ‘s opening of whole new areas of work, inquiry, creativity and innovation.
Very interesting and insightful post by Gary. While I agree that ChatGPT 3 is error-prone, carries misinformation, does not offer source, it is worth remembering Clayton Christenson's Disruptive Innovation Model- While the incumbents are busy serving high quality products to their more demanding and profitable customer base, they tend to ignore the needs of others. With ChatGPTs ability to provide a cogent response (error-prone) , the customer base might tend to prioritize convenience(effortless) over effortful selection within links offered by the incumbent. In any case, arrival of ChatGPT is a watershed moment..